kanzure

It gets bad when the subculture inhabitants get confused and adopt the social norms of the passer bys instead of remembering the importance of well-kept gardens. And then they start to introduce "policy".. which is just a backdoor for attackers to destroy your well-kept garden by arguing policy (as if that was what the subculture was about!).

Comment · 2 months ago

These models seem to be bad at writing prose or text. Many of the sentence structures seem to be unvaried.

Comment · 2 months ago

Yep. Same here. I frequently tell agents things like "answer using only a single sentence" and "write no more than 10 words". They are excellent at writing code, so have them write code (and not English prose). Besides, most of the time we want them to make reusable software that doesn't require users (or future agents) to read too much text. Software should generally just work and do the obvious thing, without needing verbose explanation.

Comment · 2 months ago

Quality control and testing is absolutely essential, yeah. It's hard work and the plan shouldn't be misconstrued as "rewrite the whole genome of an embryo today and go for the first birth immediately" because that would be stupid when we know there is other work to be done. It shouldn't surprise ANYONE that making significant progress against cancer, aging, or other issues, would take lots of work and not doing the immediate obvious broken thing.

Comment · 2 months ago

It depends on what you consider to be speculative. It's certainly not a product available to the mass consumer market, if that's what you mean. But that's also kind of an uninteresting fact, right?

"Synthetic genomes unveil the effects of synonymous recoding" https://www.biorxiv.org/content/10.1101/2024.06.16.599206v1

"Probing the limits of genetic recoding using multi-omics-guided evolution" https://www.nature.com/articles/s41467-026-74300-9

"Escherichia coli with a 57-codon genetic code" https://www.science.org/doi/abs/10.1126/science.ady4368

Comment · 2 months ago

Well, first, I would have to check and see if the synoynmous codons are important for virus reproduction. Second, this technique requires (AFAIK) lots of cell line viability testing and lots of tweaking. I don't think it's as simple as code swapping all the codons and then all the biology magically works the right way, biology is rarely that simple. Requires effort.

Comment · 2 months ago

> We will cure virus.

It is possible to prevent most viruses by recoding the human genome (requires embryo engineering or pre-embryo engineering) to use alternate codons that existing natty viruses don't use. This wouldn't prevent synthetic viruses from being developed that use the alternate codons, of course. See (1) https://diyhpl.us/wiki/transcripts/hgp-write/2016-05-10/ultr... and (2) https://diyhpl.us/wiki/transcripts/hgp-write/2017-05-09/ultr... and (3) https://x.com/RokoMijic/status/1720875112541364535 for some background.

As for "curing cancer", I suspect that Anthropic or OpenAI will use their high status/reputation to launder the existing solutions (germline, p53, enhanced DNA damage repair pathways, higher tumor-suppressor-to-oncogene copy number, more robust tumor-suppressor networks, more self-anti-cancer apoptosis, CD47 suppression, etc) that society has refused to accept. They might have enough reputation to get people to look at these ideas. It won't cure people with stage 5 turbo-cancer a.k.a the walking dead, but hopefully we won't be moving the goal posts _that_ much on them.

Comment · 2 months ago

At this point, I'm basically telling models to not write any English text or prose. Only write code. They are great at writing code. Not so great at writing good English. In software projects, lengthy comments and docs are an anti-pattern: the software should instead be written to do the right expected thing so that you don't have to think about it. I don't want all these tokens polluting my context, either.